Use fbAntares to prepare flow-based simulations

a - Function presentation

The function computeFB() converts flow-based domains (e.g. historical flow-based domains of some “typical representative days”) into an Antares model with fixed PTDF. It uses the main following parameters :

• PTDF : PTDF file path
• outputName : name of output directory

PTDF is a csv file with the following columns :

• Date : Id of typical day (integer)
• Period : Hour (integer from 1 to 24), period i corresponds to the hourly time-step [i-1 ; i]
• ptdf columns, for example :
• ptdfBE
• ptdfDE
• ptdfFR
• ptdfNL
• ptdfAT
• ram : margin
• Class : class of typical day (for example WinterSe)
• idDayType : id of the typical day
• TypicalDay : is the day a typical day (optional)

A PTDF file at this format can be automatically generated by using the function writeAllTypeDays of the package fbClust. One can also set the following optionnal arguments of the function :

• dayType : subset of typical days to process
• hour : subset of hours to process
• nbFaces : number of binding constraints in the final Antares model (by default : 75).
• areaName The name of the area of your study, possible values are cwe_at (default), cwe and other. If you choose other, you have to modify the csv file of the package and use the examples to write how your area work. You can access to this file at the path “areaName/areaName.csv” in the package fbAntares. It contains four columns (area, links, country and antares) which are, for a given area (for example cwe_at), the links between the countries of this area (ex : at.de, be.fr…), the countries in this area (ex : AT, BE, FR…) and how those links are translated (ex : BE.FR|BE - FR, DE.FR|DE - FR, DE.NL|DE, BE.NL|BE, BE.DE|BE - DE,AT.DE|AT - BE) with the name of a link and the operation made to compute it. Rq : If you use the virtual flowbased area (param virtualFBarea = T), you don’t have to change the areaName parameter.
• fixFaces (data.table) If you want to have fixFaces in your modelized domains, the fixFaces can only be the maximum of import (func = “min”) or maximum of export (func = “max”)
• hubDrop list of hubs in the ptdf, with the ones which should sustracted to the others as the names of the arrays which themself contain the ones which be sustracted
• virtualFBarea logical,
• TRUE if you want to use the virtual links between area, ie if you use a virtual area named zz_flowbased, linked with all the real areas you have in your data (for example be, de, fr, nl, at). -FALSE if you use the real links (at - de, be - fr, be - nl…).

To learn more, including tuning arguments for the optimization problem, see ?computeFB for more details.

The function creates a directory with the following files :

• weight.txt : PTDF for the antares model (weight of the binding constraints - constant over time)
• second_member.txt : RAM for the antares model (second member of the binding constraints - variable over the typical days and hours)
• domainesFB.RDS : Information about the modeled domains (points, errors , accuracy of the model,…)

## Example of computeFB with cwe_at
computeFB(PTDF = system.file("/input/ptdf/ptdfraw.csv", package = "fbAntares"),
          reports = FALSE, areaName = "cwe_at", 
          hubDrop = list(NL = c("BE", "DE", "FR", "AT")),
          nbFaces = 75, dayType = 1,
          fixFaces = data.table(func = "min", zone = "BE"))

## Another example with more arguments and faces clustering only on most important hours
computeFB(PTDF = system.file("testdata/2019-07-18ptdfraw.csv",
                             package = "fbAntares"),
          reports = FALSE, areaName = "cwe_at", 
          hubDrop = list(NL = c("BE", "DE", "FR", "AT")),
          nbFaces = 75, dayType = 1, 
          clusteringHours = c(7:10, 17:19), nbLines = 50000,
          maxiter = 20, thresholdIndic = 95, 
          fixFaces = data.table(func = "min", zone = "BE"))

b - The using of the virtual area

Two different representations of flow-based can be used in Antares simulations. The historical one uses “real links” and translates the PTDF of the domain into a linear combination of exchanges. For example, PTDF of France will apply on the exchanges between France and Belgium and the exchanges between France and Germany. An alternative representation consists in creating a virtual area named zz_flowbased to which all areas are connected. In this case, this becomes the only link between each area and the other flow-based areas. It is a more pedagogic representation because it illustrates perfectly the fact that France exchanges electricity with the entire flow-based area at once and there is no differentiation between to exchanges with Belgium and those with Germany.

To use this alternative modelling, set to TRUE the parameter virtualFBarea in the function computeFB. The functions initFlowBased and the afterAntares funcitons will then adapt by themselves.

## Example of computeFB with cwe_at and virtual area
computeFB(PTDF = system.file("/input/ptdf/ptdfraw.csv", package = "fbAntares"),
          reports = FALSE, areaName = "cwe_at", 
          hubDrop = list(NL = c("BE", "DE", "FR", "AT")),
          nbFaces = 75, dayType = 1,
          fixFaces = data.table(func = "min", zone = "BE"),
          virtualFBarea = TRUE)

c - Validation of domains results

The function getMaxImportExport() returns a data.table and a csv if you want to. The result is the difference between the maximum of imports and exports between the modelized domain and the real domain for the different areas

#Generate reports from the output of computeFB
domainesFB <- readRDS(system.file("/input/model/antaresInput/domainesFB.RDS", package = "fbAntares"))

# To check on the modelization results
maxImportExport <- getMaxImportExport(domainesFB, writecsv = F)

# for a good visualisation in this vignette :
DT::datatable(maxImportExport, options = list(scrollX = TRUE))

a - Presentation of the function

The previous function - computeFB() - converts the typical flow-based days in a format suited for Antares.

The next step consists in defining, for each day of each Monte-Carlo year, which typical flow-based day will be used to bound the exchanges within the CWE-AT (or CWE) area. The function createFBTS() performs this task. Possible correlations between the flow-based shapes and external variables (e.g. load, wind power generation, solar generation) can be taken into account. Therefore, the function createFBTS() considers several inputs.

The main inputs of this function are :

• probabilityMatrix : the probability matrix carry the information on the correlations between the typical days and external variables. They are calculated with the function getProbability() of the package fbClust. The names of the columns of the probability matrix must correspond to inputs of the Antares study by respecting the following syntax area@variable (e.g. : fr@load, de@solar, be@wind). If needed, the columns can be easily renamed with the function setNamesProbabilityMatrix().

• multiplier : The probability matrix levels are usually computed on normalized data (e.g. load factor for RES generation). They can be rescale to the time series of the Antares study by using a multiplier. The multiplier is usually the installed capacity (in MW).

• calendar and firstDay : defines the calendar, i.e. the the boundaries between the different seasons (winter, summer, interseason) and the week-end days. the firstDay can be automatically identified with the function identifyFirstDays() which will look for patterns in the load time series to identify which days are saturdays and sundays.

• opts : a link toward an antares study, as returned by the function setSimulationPath() from the package antaresRead.

• outputPath : directory where the file ts.txt, containing the time series of typical flow-based days, will be written.

An example of the use of createFBTS() is given below.

# build probabilityMatrix with the function getProbability() from the package
# fbClust
# select an antaresStudy with the function setSimulationPath() from the package antaresRead

# rename columns of the probability matrix

matProb <- setNamesProbabilityMatrix(probabilityMatrix, 
                                     c("FR_load", "DE_wind", "DE_solar"),
                                     c("fr@load", "de@wind", "de@solar"))

# set installed capacity for Wind (addition of onshore and offshore) and solar
multiplier <- data.frame(variable = c("fr@load", "de@wind", "de@solar"),
                         coef = c(1, 71900, 61900))

# set Calendar
firstDay <- identifyFirstDay(opts = antaresStudy)
calendar <- system.file("calendar/calendar.csv", package = "fbAntares")

# create ts.txt in D:/model1
ts <- createFBTS(opts = antaresStudy, probabilityMatrix = probabilityMatrix,
                 multiplier = multiplier, calendar = calendar,
                 firstDay = firstDay, outputPath = "D:/model1")

b - Validation of the results

getStatsFBts() gives detailed informations about the time series generated with the previous function createFBTS(). You just have to give in input the object ts generated with createFBTS(), a calendar with the csv format (like the one available in “/calendar/” in the package fbAntares) and an output (“summary”, “yearbyyear” or “all”)

ts <- fread(system.file("testdata/antaresInput/ts.txt", package = "fbAntares"), header = T)

statsFBts <- getStatsFBts(ts, calendar = system.file("calendar/calendar.txt", 
                                        package = "fbAntares"), output = "summary")

# for visualization in the vignette :
datatable(statsFBts, options = list(scrollX = TRUE))